GEOPHYSICS,

VOL. 55, NO. 7 (JULY

1990); P. 902-913, I1 FlGS

Wavelet estimation for a multidimensional acoustic

or elastic earth

Arthur B. Weglein* and Bruce G. Secrest$

ABSTRACT

A new and general wave theoretical wavelet estimation method is derived. Knowing the seismic wavelet
is important both for processing seismic data and for
modeling the seismic response. To obtain the wavelet,
both statistical (e.g., Wiener-Levinson) and deterministic (matching surface seismic to well-log data) methods are generally used. In the marine case, a far-field
signature is often obtained with a deep-towed hydrophone. The statistical methods do not allow obtaining
the phase of the wavelet, whereas the deterministic
method obviously requires data from a well. The
deep-towed hydrophone requires that the water be
deep enough for the hydrophone to be in the far field
and in addition that the reflections from the water
bottom and structure do not corrupt the measured
wavelet. None of the methods address the source
array pattern, which is important for amplitude-versus-offset (AVO) studies.
This paper presents a method of calculating the total

INTRODUCTION

.
I
. .
In seismic exploration a man-made source ot energy
produces a wave which propagatesinto the subsurface. The
reflection data recorded on the surface depend on (1) the
properties of the earths structure (2) the energy source and
recording system.
The purpose of seismic exploration is to extract information about the subsurface from these data. Consequently, it
is important to attempt to identify and remove the effects of
the source characteristics from this reflected energy. We
suggestthat a wavelet estimation method be applied which is
theoretically consistent with the processto be applied to the
reflection data. That is, for acoustic processing(e.g., acous-

wavelet, including the phase and source-array pattern.

When the source locations are specified, the method
predicts the source spectrum. When the source is
completely unknown (discrete and/or continuously
distributed) the method predicts the wavefield due to
this source. The method is in principle exact and yet
no information about the properties of the earth is
required. In addition, the theory allows either an
acoustic wavelet (marine) or an elastic wavelet (land),
so the wavelet is consistent with the earth model to be
used in processing the data. To accomplish this, the
method requires a new data collection procedure. It
requires that the field and its normal derivative be
measured on a surface. The procedure allows the
multidimensional earth properties to be arbitrary and
acts like a filter to eliminate the scattered energy from
the wavelet calculation. The elastic wavelet estimation
theory applied in this method may allow a true land
wavelet to be obtained. Along with the derivation of
the procedure, we present analytic and synthetic examples.

tic migration) an acoustic wavelet would be appropriate,

whereas an elastic wavelet would be appropriate for elastic
wave-equation data processing.
It has been shown [Loveridge et al. (1984)] that for
amplitude-versus-offset (AVO) studies, the source array
pattern can be important. We show a method of obtaining
this array pattern.
The recently published one-dimensional (1-D) acoustic
wavelet estimation method of Loewenthal et al. (1985)
assumes (1) that the medium above the receivers is known
and (2) that both the field and the normal derivative are
measured. Our method makes analogous assumptions for
multidimensional acoustic and elastic media. In a related

Manuscript received by the Editor March 6, 1989;revised manuscriptreceived December 4, 1989.

*Formerly BP Exploration, 5151 San Felipe, P.O. Box 4587, Houston, TX; presently ARC0 Oil and Gas Company, 2300West Plano Parkway,
Plano, TX 75075.
SBP Exploration, 5151 San Felipe, P.O. Box 4587, Houston, TX.
0 1990Society of Exploration Geophysicists. All rights reserved.

903

Wavelet Estimation
paper, Sonneland et al. (1986) use the 1-D acoustic wave
equation with two independent vertical measurementsof the
pressure to perform a combined designature-dereverberation. Hargreaves (1984) has presenteda wave-field extrapolation method for source signature identification. Although
his method produced favorable results in some cases, e.g.,
deep water, he states a problem resulting from corruption
due to the scatteredfield. The method in this paper directly
addressesthis problem, automatically filtering the scattered
field from the wavelet calculation.
Our purposeis to present a new general method of source
signature identification which also requires two separate
field measurements.It is applicablefor an arbitrary inhomogeneousmultidimensional acousticor elastic earth. Furthermore, no information about the propertiesof the subsurface
is required. Each shot record producesan effective acoustic
or elastic wavelet for that particular shot record. The effective wavelet can vary from one shot record to the next. The
wavelet obtained is the source wavelet or driving function
being imparted to the medium. In the second section, we
present the general wavelet estimation procedure and show
that the method relies only on the incident wave, effectively
filtering the scatteredenergy from the integral. We illustrate
this procedure with an analytic example. The marine exploration environment is treated next. We show an application
to a distributed source and the source array pattern, followed by an application of the method for a multidimensional elastic problem. Finally, we give numerical examples
which addressthe issuesof time and spatial sampling, finite
aperture, approximations to the field derivative, and finding
the source array pattern.
The procedure (for either the acoustic or elastic model)
derives and from a comparison of the Lippmann-Schwinger
equation and Greens theorem. The former originates in
scatteringtheory and the latter in boundary-value problems.
These two equationswere compared,for a different purpose,
in Weglein and Silvia (1981) and Silvia and Weglein (1981).

w2

( )
V2 + 2

P = 2

(2)

We now proceedto derive two integral equationsfor p. One

is the Lippmann-Schwingerintegral equation and the other is
derived from Greens secondidentity. These two equations
will lead to the desiredrelationship between the soughtafter
A(w) and the measurementsof p on the surface z = 0.
Formally inverting the operator V2 + w2/ci = La = Cc
in equation (2), we have the operator relationship

lo* _ _

P=GoTaP+AGo
co

or, equivalently,
P(r, rs9 01 = AblG~(r,

rs, 4

co=

G&, r, w) _r a(r)P(r,

rs, O) dr,

(3)

CO

where
G&,

r, w) = S(r - r)

with outgoing wave-boundary conditions.

Equation (3) is the Lippmann-Schwinger integral equation
and is valid for all r.
A secondintegral equation for P is derived from Greens
secondidentity:
(AV2B - BVA)

(AVB - BVA)

d3r =
I

. n ds,

where S is the surface which encloses the volume V. Let

A = P and B = Go in Greens second identity:

METHOD FOR ACOUSTICWAVELET

Consider a point source at the spatial position r,. The
constant-density acoustic wave equation for the pressure
field P due to a sourceA(t) at r,? is

[h-,

rs, w)V*Go(r,

r, W)

- Go(r, r, w)Vt2p(r,

(y&g)

09 + A(w)s(r - r,).

rs, w)] dr

P(r, rsr t) = A(t)G(r - r,).

=

[f%,

rs, w)VGo(r,

r, w)

Taking the Fourier transform in time gives

iJJ2 _

( >
V2 + c(r)

- Go(r, r, o)VP(r,

P(r, rs, w) = A(w)s(r - r,).

In this context the wavelet estimation problem is to determine A(w). Characterize the velocity configuration c(r) in
terms of a reference value cn and a variation in the index of
refraction (Y:
1

c(rj

I
= - [I - ff(r)].
ci

Equation (1) can then be rewritten as

rs, w)] - n ds,

(4)

(1)

where S is the surfaceillustrated in Figure 1. rs is located on

or above the z = 0 boundary of the surfaceS. We choosethe
convention for the delta function to be

6(r - a)f(r) dr =f(i),

P is strictly within the volume. For amoutside the

volume, or on the boundary, this integral vanishes.
Substituting the equations

where

Weglein and Secrest

904
V P
(rl,

Lo2
r,5 ,

0) =

P(r,

-1

I
r,s, 0)

A(w) = Go(r, rs, co)

(0

+ 1 a(r)P(r,
co

r,,

, 0) + A(w)S(r

*I

- r,)

[I%-,

and

- COW.
2

Vf2Go(r,

r, w) = -q

Go(r,

r, w) + S(r - r)

co

into equation (4) and taking r within the volume, we obtain

P(r,

rs, 0) =

Go(r,

r,

co)

co

[%-, rs,

- G(r,

r,, , o) dr

a(r)B(r,

r, 01

o)VGo(r,

r, w)VP(r,

rr,

o)] * n ds.

(5)

Thus, for r within the volume we have two expressions for

p. These two integral equations [equations (3) and (5)] were
studied in by Weglein and Silvia (1981).
If the support of a is within the volume V, then

w)

[f%Go
a(r)&,

r, w)

r, w)Vp(r,

r,,

co)] * n ds. (6)

The upper boundary of the volume V is defined by the

measurement surface z = 0. Therefore the support of (Y
within V means that the reference medium must only agree
with the actual medium above the receiver plane. The
numerator and denominator can be evaluated at any point
inside the volume. In principle, one evaluation point will
suffice to determine /i(w). However, in practice N such
evaluations of the numerator and denominator will give N
independent estimates of A(w) for each source location. A
statistical weighting of these estimates could then be used to
evaluate the wavelet optimally. Consequently, in practice
the method would involve a combination of deterministic
and statistical procedures. We show later that this redundancy can be used to solve for the source array pattern.
The total field P and its normal derivative VP . n depend
on both the wavelet and the subsurface properties. However. we have shown that

Go(r, r,

r,, w)VGo(r.

- GoVP]

- n ds

rs, O) dr

c;

w2
Go(r,

r,

co) 1 a(r)P(r,
C6

r,, , w) dr.

For points r within the volume and the support of (Y within

the volume V, equations (3) and (5) give two expressions for
the field in the reference medium. From these we obtain

depends only on the wavelet. The major conclusion is that

measurement of ls and VP . n on the surface of the earth is
sufficient to determine the source wavelet without requiring
any information about the subsurface properties. We now
show that the above integral depends only on the incident
field by showing that the integral for the scattered wave is
zero. Thus, since
Go(r, r, w) = 6(r - r)
and

- Go(r,

r, o)V&,

r,r, w)]

. n ds

or

EARTH

FIG. 2. The marine exploration geometry.

AIR
WATER

FIG. 1. Subsurface volume V.

Fto. 3. Background medium for marine case.

905

Wavelet Estimation

_!-

Cci

Gj

I
=
[l - h6(x -X0)],
c(r)
co

we have, by substituting into Greens second identity,

[P,VG,,

- GuVP,]

i.e., (Y = hS(x - x0).

We evaluate equation (6) as

- n ds

/ s

[pVGO - GOVP]
W2
Go 7 a(r)P

= p, I

dr = p,, - P, = 0,

- n ds.

The Greens function Go is given by

CiJ

and thus the only contribution to the integral in equation (6)

is the incident wave.
We illustrate this method with 1-D and multidimensional
examples.

I
Go

One-dimensional example-Localized

11

eikix

2ik
and satisfies
d2Go
dx2 + k2Go = 6(x -x),

WAVELET ESTIMATION
We now demonstrate the method with the analytic example of a point scatterer. Numerical examples of the method
will be given later.

where k = o/co and cg is the reference velocity.

For an impulsive source at x,, the incident field is

inhomogeneity
2ik

We illustrate this method for the I-D model

OFFSET

(FT)

FIG. 4. Shot gather: point source, incident wave.

(8)

906

Weglein and Secrest

The total wave field P satisfies

and P(.u, x,, O) from equation (1 la) becomes

d2P
---Q

k2(1 - oi)P = A(o)S(x- - X,).

(9)

The problem is to determine the wavelet A(w) from the

boundary measurements of P. The Lippmann-Schwinger
equation, valid for all x, is

--

A A(w)P

iklx- roieikl.ro

-xti
(I lb)

ikX
I+7

or

The total wave field p(x, x,, w) is a function of the wavelet

A(w) and the medium.
The secondintegral representation, derived from Greens
second identity, is
1
+-

2ik

(~k~.ika(x)P(x.

p-----

x,~, w) dx.

d"P

d'G,,
dx2

Go

dx2

dx

For the localized scatterer, (Y = hF(x - x0), the total wave

field is
= 1; (+G,,

g),

(12)

where a < x,) < h. Using the differential equations (8) and (9)
in the left-hand side of equation (12) leads to
(1 la)

To find P(n-,,, x,\, w), set x = x0 in the last equation:

Ih

[P(x,

x,, , w)S(x - x)

k2a(x)P(x, x,~, o)Go(x,

- Go(x, x, w)&o)S(x - x,~)]dx'

or

_
I(
b

2ik
&I,

x, co)

dGo

di,

pdx-G,dx

x,s, w) =

ikh

For CI< x < b and x,~< a, this last expression is

l+y

FIG. 5. Wavelet reconstruction for shot gather in Figure 4

with analytic derivative.

FK. 6. Wavelet reconstruction for shot gather in Figure 4

with Eulerian approximation to derivative.

Wavelet Estimation

k=

Go(x, x, w)a(x)li(x,

xs, w) dx

907

Go(x,

x,

w)k=a(x)P(x,

o) dx

x,,

a
^a
=

dGo(x,

&,

II

x, >0)

x,

o)k*a(x)P(x,

x,~. w) dx,

and for x within the interval (a, b), it follows from equations
(lla) and (13) that

dx'

- Go(x, x, 01

Go(x,
s --53

x, 0)

dP(x, x, > 01

dx'

and, solving for P(x, xS, o),

dGo(x,
P(x,

WI

x,,

xs,

CO) = k=

- Go(x,
G,,(x,

x,

w)a(x)P(x,

I,

co)

dx

x, , w)

d&,
b
p(x,

x,

w)

xs, w) dx

dx

and
b

Ii

&,

dGo(x,
x,

1 0)

x,

$0)

w)

dx

- Gotx, x, w)

drj(x,

dGo

P(x,

x,,

w) dx

(13)

- Go(x,
/gklx

For this example, the support of a(x) is at x0, thus

2ik

OFFSET (FT)

FIG. 7. Shot gather: point source, plane reflector.

x, j

x,

w)

x5, w)
dx

(14)

Weglein and Secrest

908

Removing

This is obviously the 1-D reduction for our wavelet detcrmination scheme, equation (6); i.e.,

@Go

the absolute value signs with a

< a, we arrive at

< x

< b,

u i x0 < b, andx,

- GoVf) * n ds

I
k(o)

2ik

Cl,

which was to be shown. Notice that although P and #Iian

depend on the medium through co, h, and x,,, the integral

To complete the example, WC show that the right-hand

side of equation (14) just reduces to k(w). Thus we evaluate
the right-hand side of equation (14) for the case of a localized
scatterer. We need the Greens function.

is only a function of the source signature.

and its derivative,
dGo
-=

&V mxi
~

dx

From equation (lib),

SGN

WAVELET
ESTIMATION
IN A SPATIALLY
VARIANT REFERENCE MEDIUM

(x -x).

As stated in the Introduction, it is apparent that the

method for obtaining the wavelet just described will hold for
more general cases, in particular for a spatially variant
reference medium. In a similar manner, an expression for the
wavelet can be derived for the case where the acoustic
velocity is characterized in terms of a spatially variant
reference velocity. This spatially variant reference velocity
will be required for the marine wavelet estimation. As
before, the reference medium must agree with the actual
medium above the receivers.

the wave field is

cl --x,1

and its derivative is

SGN

(x -x,)

Special case-Marine

ik SGN

wavelet estimation

The marine exploration configuration is illustrated in Figure 2. The background medium for this application will
consist of two homogeneous half-spaces illustrated in Figure
3.

(x -x,,)

We now evaluate equation (14) to obtain

b

- Go(x, x,

k(w)

d& , x.7,

dGo(x, x, ~1

&, xs WI

dx

p+

w)

~1

dx

--x., j
2ik

and thus, upon substituting for P, G, and their derivatives,

r

eikb --xs
~--

2ik

eiklY -rii~elk~il

--x5

eikl\ -x/
~

SGN (x -x)

2ik

-~

~
2ik

SGN

(x -x,~)

- ;

ik SGN

iti
l+2

(x -x0)

II
.

909

Wavelet Estimation
where P satisfies the differential equation

For this geometry, G,, is given by the equation

VP(r,

I-~,

w) +

[l - a(r)]P(r,

r,T, w)

Cl)

=,4(w) i

6(r -

r.5j)

;=I

and is th-e total fieldsdue tomthe ;r distributed point sourcesi

However, if the localized point sources are each different,
we can use the fact that r is arbitrary to obtain the system of
equations

and dGo/dz,7 is given by

dGo
-=
dz,

I_:

B(r, r,, . . . ry,) , 0)

J&h,

r, ~1
an

dP(r, r,s, . . . rs, , 0)

ii&

- GA-,,

r, 0)

ds,

an

06)

i=l,...m.

This system of equations can be solved in a least-squares

sense for the k,(w).
Alternatively, the radiation pattern from a single effective
point source could be determined by assuming that k(o) is
also a function of the radius r, i.e., A(w)6(r - r,?) becomes
the more general k(w, r),

k," SGN (z, - z)

co
k(o,
I --(c

ZZ fib,
Is

Si( J(;)2-k;(z+z,y))
x (-id(,)I-k:)}

SOURCE

If the source consists of a sequence of localized point

sources at r,v,, rs,, . . _ , r7 , each with same time dependence, then the wavelet A(;) is given by
k(w) =

%(r,r,co)
-G~(r,r,w)
an

r, o)VP(r,

r, w)

rs, co)] * n ds,

(17)

and thus the relative strength of the source field can be

obtained by evaluating equation (17) for various r on a
constant radius about the source.

where ct , c2 are the acoustic velocities of water and air,

respectively, and A = (l/c:) - (l/c;).

&-,r,,w)

r, co)dr

r,, w)VG&,

- G,,(r,

dk,,

DISTRIBUTED

r)G,,(r,

@r, r,, ~1
an

ds
?

(15)

ELASTIC

WAVELET

ESTIMATION

METHOD

Below WC outline the generalization of the wavelet estimation method for elastic waves. As in the acoustic case, a
comparison of the Lippmann-Schwinger
equation and
Greens theorem is used to determine the wavelet.
To derive the Greens theorem for the elastic case, a rank
two Greens displacement tensor so is_ defined and in
addition a rank three Greens stress tensor $ is defined. In an
analogous manner to the scalar case, a Greens identity is
obtained relating the displacement u(x, t) to a volume
integral involving a body force and an integral over the
enclosing surface which involves the displacement and its
derivatives through the traction t. This Greens theorem
result [e.g., Pao and Varatharajulu (1976), equation (19)] for
elastic waves is given by

910

Weglein and Secrest

pf(x)

* i&(x,

x)

dx

{t(x)

* &(x,

x) - u(x)

* [n - $(x1x)]} cf.5

zz

44

XEV

x @

v,

where u(x, t) is the displacement vector, t = n .f, where t

is the traction on the surface, T is the stress tensor, n is the
unit vector normal to the surface, f is the body force=per
unit mass, CO is the Greens displacement tensor and 7 is
the third-rank Greens stress tensor.
Let GO be the response in a homogeneous medium and f
the passive sources (the scattering centers) of the scattered
displacement field originating from the inhomogeneities in
the medium. The physical source is outside the volume V. In
Pao and Varatharajulu, f represents the body force per unit
mass and is inside the volume. They were considering an
active physical source inside the homogeneous medium. We

FIG.

OFFSET

0 0 000

9. Wavelet reconstruction for shot gathers in Figures 7

and 8.

(FT)

-2700
llllllllllllllllllllllIllllll,PilIIllIllllllllllllllllllIIlll

0.100

0.200

0.300

0.400

0 - 500

0.600
FIG. 8. Shot gather: point source, partial plane reflector.

+2700

911

Wavelet Estimation

are considering a physical source outside an inhomogeneous

medium with passive scattering sources inside V.
Let umn(x, x) be the displacement at x in the n-axis
direction due to a source at x in the m-axis direction. The
source is A,(w)s(x - x)II where II is the unit dyadic. In a
manner similar to the acoustic case, the wavelet A,(o) can
be determined from the Lippmann-Schwinger equation for
elastic waves by dividing the surface term in equation (20) of
Pao and Varatharajulu (the mnth component),
{(Go - n)(V

* u) - (V * (&)(u * n)} ds

&-, r, , 0)

dG~(r', r, ~1
an

- PCS {& - (n
/ s

u)

(V

Go)

(n

u)}

ds
n

by G,,.
NUMERICAL

EXAMPLES

Numerical examplesare now used to investigatethe effects

of time and spatialsampling,finite aperture, approximationsto
the derivative of the field and the sourcearray pattern. These
examples are provided by solving a 2-D acoustic forward
problem with a free surface, and a point source at xs. The
wavelet estimation equation for this case is then

aP

- Gob, r, w) ;

(r, rs, 0)

ds

Go@, rs, ~1

OFFSET (FT)
-2700

0 m000

0.

+2700

Illllllllllllllllllllllllllll19llllllllllllllllllllllllllllll

I00

0.200

0.300
F

0.400

0.500

0.600
FIG. 10. Shot gather: distributed source, plane reflector.

Wavelet Estimation
source distribution. For the latter, no information about the
nature of the source was required.

ACKNOWLEDGMENTS
The authors wish to thank BP Exploration

for permission

to publish this paper. They would also like to thank ARC0

management for encouraging and supporting this research.

REFERENCES
Hargreaves, N. D., 1984, Far-field signatures by wavefield extrapelation: Presented at 46th annual EAEG meeting, London.
Loewenthal, D., Lee, S. S., and Gardner, G. H. F., 1985, Deter-

913

ministicestimationof a wavelet usingimpedancetype technique:

Geophys. Prosp., 33, 956-969.
Loveridge. M. M., Parkes, G. E., Hatton, L., and Worthington, M.
H.. 1984. Effects of marine source array directivitv on seismic
data and source signature deconvolutionl First Breik, July,
Pao, Y. H., and Varatharajulu, V., 1976, Huygens principle,
radiation conditions, and integral formulas for the scattering of
elastic waves: J. Acoust. Sot. Am., 59, 1361-1371.
Secrest, B. G.? and MacBain, J. A.. 1990, Source identification in
remote sensmg problems: Proc. of SIAM Conf. on mathematical
and computational issues in geophysical fluid and solid mechanits, Houston, to appear.
Silvia, M. T.. and Weglein, A. B., 1981, A method for obtaining a
near-field inverse scattermg solution to the acoustic wave equation: J. Acoust. Sot. Am., 69, 478-482.
Sonneland. L., Berg, L. E., Haugen, A., Fotland, B., and Vestby,
J., 1986, 2-D deghosting using vertical receiver arrays: 56th Ann.
Internat. Mtg.. Sot. Expl. Geophys., Expanded Abstracts, 516519
I *Weglem,
A. B., and Silvia, M. T., 1981, A scattering theory
approach to the identification of the Helmoltz equation: A nearfield solution: J. Acoust. Sot. Am., 69, 483-488.